Taking on a tech breakthrough: Machine learning without the learning curve
Risk management professionals are voracious data consumers. They seek better intelligence to develop market insights and evaluate risk. They want to confirm marketing, acquisition and account management strategies. They must anticipate and mitigate threats while they prepare to decide in real time. They gather as much information as possible about what their organizations can do, their customers’ needs and the competition’s weaknesses.
But their jobs aren’t easy. Credit decisions, risk assessment models and marketing forecasts require better, faster, current data. Large data sets increase data’s power because it allows them to compare and contrast existing behavior with historical results across a broader pool of variables. Access to more data —internal and external—ultimately produces better outcomes. For analysis and research, data analysts and risk managers need on-demand access to ever-larger, high-quality, depersonalized, structured data.
With vast volumes of data to consider, organizations typically tap big data analytics to help assess nearly every contingency. Thanks to recent advances, the latest in machine learning and predictive analytics is available to a broader market. These advancements allow risk managers from financial institutions of all sizes to incorporate machine learning models into big data processing systems—and power fresh opportunities.
Reconsidering the comfort zone
Big data capabilities such as machine learning are important because many risk managers stay within their comfort zone, relying solely on their own internal data for predictive analysis rather than trusting outside sources. Internal data is apt to be more fully vetted, better understood and managed. But growth into new markets and strategy deployment require venturing beyond familiar territory. The best opportunities come from adding new data and better tools to the analytics toolbox.
Leveraging external data can be difficult, expensive and time-consuming. Even the purchasing process itself can prove slow and treacherous. Evaluating the most useful data sets, getting budget approvals, and uploading and filtering the information take time. Optimistically, this may represent a three-month ordeal; unfortunately by the time the data arrives, the situation may have changed.
Regulatory and privacy compliance requirements further hinder risk managers from delivering new data sets to analyze. Combining demographic and marketing information with credit data can provide tremendous analytical insight. But using credit data means increased regulatory burdens with more potential for misuse; this could foist new regulatory implications onto the marketing data.
When an organization expands its sights beyond its own data and taps external data stores, the value of analytics increases dramatically. Yet large external data sets take additional time to acquire, are tougher to process and require substantial resource investments. Big data infrastructures place large demands on servers and programmers. An effective environment that handles petabytes of data might demand more than 100 servers, something only the largest organizations can justify. This doesn’t even begin to speak to the personnel needed to manage the newly acquired infrastructure.
Bringing machine learning to data analytics
To streamline this process, cloud-based distributed solutions are now an option: integrating machine learning technology in a hosted distributed environment so it can handle massive data sets with advanced modeling to predict the schema of incoming data sets.
It’s critical to move beyond static libraries of information and see the benefit of combining clients’ internal data with vast data repositories that include access to historical credit data, full file tradeline data, public records, attributes and consumer credit scores. And when this data doesn’t include any personally identifiable information, it ensures regulatory compliance.
Making information analysis easily accessible also creates distinct competitive advantages. Identifying shifts in markets, changes in regulations or unexpected demand allows for quick course corrections. Tightening the analytic life cycle permits organizations to reach new markets and quickly respond to competitor moves.
It is of course imperative that organizations have access to their own data. However, when you combine it with a massive historical national file, companies can now benchmark their data against themselves and competitors. They can contrast specific product lines, understand why scores may shift and how that may impact existing lines of business—and as a result develop strategy modifications on the fly.
Big data’s big questions
An analytical sandbox provides ongoing snapshots of data from a national file, allowing risk analysts and line-of-business managers to run a broad range of “what if” scenarios. There is no shortage of big data questions:
- Does a regional shift in credit score portend a national trend?
- How might a spike in overall debt impact collection projections for a home equity line of credit?
- If a national retailer closes 140 stores, what opportunities could that create to target its affinity card holders?
- “We’re doing great in this product area but are we the top provider?”
- What is a new product’s effect on the overall portfolio?
Big data analytics answers some of the hardest questions and exposes opportunities before they melt into missed opportunities. Risk managers often measure success by how quickly they react to the unexpected. They must make critical decisions intuitively and without delay. Fortunately hosted alternatives bring the power of machine learning to a wider range of business challenges. For in the final analysis, big data needs powerful tools to have a huge impact.
Want more Banking Strategies? Sign up for our free newsletter!
Vijay Mehta is the chief innovation officer for Experian’s Consumer Information Services.